Items in computer-based mathematics assessment are usually randomly selected from a large item pool. This method of assembling tests cannot cater for individual needs. In this study, stratified design is proposed and examined. Results indicate that stratification can improve measurement accuracy for both low achievers and high achievers, provided that appropriate difficulty levels of items are used. However stratification can also lead to great loss in accuracy, when mismatch of difficulty level of items and examinee ability occurs.
|Publication status||Published - 2003|
CitationLeung, C.-K. (2003). Can stratification lead to improvement of computer-based mathematics assessment? In K. T. Lee & K. Mitchell (Eds.), International Conference on Computers in Education 2003: A conference of the Asia-Pacific Chapter of the Association for the Advancement of Computing in Education (AACE): The "second wave" of ICT in education: From facilitating teaching and learning to engendering education reform: Proceedings [CD-ROM] (pp. 893-898). Norfolk, VA: Association for the Advance of Computing in Education.
- Educational Evaluation